carbon-design-system

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFE
Full Analysis
  • [Prompt Injection] (SAFE): No instructions were found that attempt to bypass safety filters or override the agent's core instructions. The content is strictly limited to UI design guidance.
  • [Data Exposure & Exfiltration] (SAFE): No sensitive file paths, hardcoded credentials, or suspicious network operations were identified. The network references are limited to legitimate CDNs (unpkg.com, Google Fonts).
  • [Obfuscation] (SAFE): No encoded text, zero-width characters, or hidden unicode markers were found in the skill or reference files.
  • [Unverifiable Dependencies & Remote Code Execution] (SAFE): The skill references well-known packages (@carbon/react, sass) and assets from trusted CDNs. No piped remote execution or dynamic code evaluation patterns were detected.
  • [Privilege Escalation] (SAFE): No commands for elevating system privileges (e.g., sudo, chmod) or modifying system configurations were present.
  • [Persistence Mechanisms] (SAFE): No attempts to create scheduled tasks, modify shell profiles, or install background services were detected.
  • [Metadata Poisoning] (SAFE): The metadata (name, description) accurately reflects the technical contents of the skill without deceptive instructions.
  • [Indirect Prompt Injection] (SAFE): The skill does not demonstrate a surface for ingesting untrusted external data into the agent's prompt context beyond its static design references.
  • [Dynamic Execution] (SAFE): No use of eval, exec, or runtime compilation of generated source code was found.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:43 PM